The AIShelf Ecosystem

The AIShelf Service is what makes your team's resources available everywhere. Run it once, and every client - your VS Code extension, MCP server, and browser extension - reads from the same local store. You don't copy-paste resources between tools; they're all in sync because they all connect here.

What it does

• Keeps local copies of your team's registries so everything loads fast and works offline - stored in ~/.aishelf/
• Sign in to GitHub once - every client you use gets access automatically, credentials stored in ~/.aishelf/
• Syncs registries with GitHub - pull, commit, push - on request
• Runs on http://localhost:5314 - every client on your machine connects to it
• Keeps your local resources secure - only trusted clients on your machine can connect

Key design principle

No cloud dependency for local operations

Once your registries are cloned, the service works entirely offline - browse and read resources without an internet connection. GitHub is only contacted when you choose to sync or push changes. Your credentials stay on your machine in ~/.aishelf/ - not in any cloud service.

Health check

curl http://localhost:5314/health
# Expected: {"status":"ok"}

The VS Code extension is the primary way to interact with AIShelf from your IDE. It provides a sidebar panel for browsing registries, an inline preview for resources, and tools for creating and pushing content back to the registry.

What you can do

• Browse all connected registries in a tree view from the IDE sidebar
• Preview resources inline without switching context
• Copy any resource to the clipboard with one click
• Insert resources directly into AI tools - Cursor's .cursorrules, Windsurf custom rules, and more
• Create and edit resources as local drafts, then push to the registry when ready
• Sync registries to pull the latest team updates

Compatible IDEs

Works with any VS Code-based editor - VS Code, Cursor, Windsurf, and others (version 1.105.0+).

MCP ServerEarly Preview

The AIShelf MCP server exposes your registry resources to AI models via the Model Context Protocol. Once configured, AI assistants can discover and apply your team's workflows, rules, prompts, and skills automatically - without any manual copy-paste.

Compatible clients

Any MCP-capable AI client works: Claude, Cursor, Windsurf, and others that support the Model Context Protocol.

Two tools

• list_resources - returns an index of all resources across your connected registries, with enough metadata for the AI to identify what to fetch
• get_resource - fetches the full content of a specific resource by its path

Configuration

Add this to your IDE's MCP configuration file:

{
  "mcpServers": {
    "aishelf": {
      "command": "npx",
      "args": ["-y", "@aishelf/mcp-server"]
    }
  }
}

Browser ExtensionComing Soon

The browser extension will bring AIShelf to any web-based AI tool - ChatGPT, Claude.ai, Gemini, and others. Browse your registry and insert resources directly from the browser, using the same service hub running on your machine.

Claude Code Plugins & SkillsComing Soon

An upcoming integration that will bring AIShelf resources directly into Claude Code sessions. The goal is to let Claude Code automatically discover and apply team-maintained workflows, rules, and skills from your AIShelf registries - without any manual setup per project.